Conda recipe examples for Deadline Cloud

The conda_recipes directory on the GitHub website contains sample tools for creating an Amazon S3 conda channel and building packages into it. The recipes use rattler-build on the prefix.dev website or, in some cases, the older conda-build on the conda website.

Each recipe directory includes a deadline-cloud.yaml metadata file that lists the conda platforms the recipe builds for and configures how the recipe is submitted to Deadline Cloud. The submit-package-job command in the conda_recipes directory submits a recipe to a queue whose name starts with Package and uses the queue's job attachments bucket as the conda channel.

For setup, see Create a conda channel using S3. For the underlying packaging concepts, see Create a conda package for an application or plugin.

To submit a Blender 4.5 build job, run the following from the conda_recipes directory of the samples repository:

./submit-package-job blender-4.5

To submit a build for both Linux and Windows:

./submit-package-job blender-4.5 -p linux-64 -p win-64

To submit a build for every platform listed in deadline-cloud.yaml:

./submit-package-job blender-4.5 --all-platforms

The conda_build_linux_package job bundle on the GitHub website is the Open Job Description template that submit-package-job submits. It runs conda-build or rattler-build on a worker, takes a recipe directory and optional source archives as input, and writes the built package to your queue's conda channel.

Source archives that recipes pull in by URL go in the archive_files directory on the GitHub website. Recipes that require an installer or redistributable archive you have to download manually (such as Maya, Cinema 4D, or Nuke) read from this directory at build time.

The following recipes on the GitHub website package the Deadline Cloud client libraries themselves so that jobs can call Deadline Cloud APIs from inside a queue:

The following sections describe the recipe families available in the samples repository.